Spatiotemporal dynamics of leaf transpiration quantified with time-series thermal imaging

Abstract

Accurately capturing the spatiotemporal dynamics of transpiration from sub-leaf to ecosystem scales remains a key challenge in eco-physiology and hydrology as typical methods face a trade-off between spatial coverage and temporal resolution. Here, we developed a new scalable, semi-automated method to produce highly precise estimates of water and energy fluxes and applied it to single leaves. High-resolution thermal infrared (TIR) images and paired colour photographs of excised soybean leaves were captured at 15 s intervals until wilting, automatically registered and segmented, and used as input for transient energy balance models to estimate latent heat flux (transpiration) at a temporal resolution of one second. Three approaches to estimating leaf boundary layer conductance to heat (gHa) and sensible heat flux were compared, two of which did not require the use of any dry or wet reference surface. The accuracy of water loss modeled using average leaf temperature was also compared to models retaining pixel-scale temperature heterogeneity at a spatial resolution of 0.326 mm2. Cumulative leaf water-losses modeled using average leaf temperature closely matched gravimetric measurements (r2 = 0.95) and pixel-scale models identified striking spatiotemporal patterns of water loss at the sub-leaf scale. Different methods of estimating gHa did not significantly alter model results. Use of leaf energy balance models with time series thermal images to quantify transient transpiration fluxes was able to accurately resolve 1-s time-varying leaf water loss in outdoor conditions, did not require any reference surfaces, and also produced data on the characteristic length scales of heterogeneous sub-leaf response. Given the ability to omit reference surfaces and retain accuracy, this approach also has the potential to be scaled-up to quantify energy fluxes in more complex plant canopies.